millenniumwaraigisfandomcom-20200223-history
Talk:Holy Warrior's Challenge/@comment-30176423-20161116032002/@comment-30176423-20161117033236
@Lzlis "What experimental data have you collected? Claiming that the math is on your side is one thing, but if it so easy, then why not do it and show us rather than making some vague claim?" Just read the comments for the past events (more specifically Belinda and Lucille). All the information is there. One of my posts was even a direct reply to you about getting 10 Lucille drops in 11 sorties. And of course, none of the runs on this map had been filtered out (BTW, 11 runs with 7SC cost takes more than 3 days, so it's difficult to collect more data). Also please pay attention to the reports of the other people. If you really read up on the math (the binomial distribution, statistical power analysis, checking if a coin is fair, p-value and other things) then you will start noticing oddities too, rather than assuming that such results still fit your model reasonably well. For example, have a look here (the original poster even did not see anything abnormal with his results). @Lzlis "There is no way you can ensure against the possibility of a correlation between someone's decision to report the data and the kind of data that was reported. (My own attempt to fight this is asking people to follow the rule "Before starting the mission, decide whether or not you will record the result" which of course only works if the result is uncorrelated from any foreknowledge the user might have -- not true in your model.)" What is not true? Do you understand that *all* results have been reported and nothing filtered out? Before starting the event we decide to watch the drop probabilities. And we also don't arbitrarily pick some convenient set of results from the middle of the event. As for the Belinda event. I already mentioned that I initially observed a consistent drop rate around ~50%, which spanned across sevaral days and got at least 10 copies of Belinda during it. Then something happened and I got around 5 or 6 runs without any drops in a row. This was very unusual, considering that this wiki was claiming that the drop rate was supposed to be 8%. That's why I decided that it was worth reporting and registered here. Then I tried to do various changes in my team setup and in the deployment order until the drop rate seemingly recovered. I don't know if it was the result of my actions, or just a temporary glitch. The chance of 5 consecutive runs without any drops at 50% drop probability is just 0.50^5 = ~3%, however allowing such gap anywhere in the midle of a sequence of 20 trials is much higher: (1 - (1 - 0.50^5) ^ (20 - 5 + 1)) = ~40%. Nevertheless, after this I also recorded videos of the next 19 runs on this map with 7 Belinda drops: http://ssvb.name/mwa-witch-girl-farming-experiment/ (you can filter out the first result, because the sequence naturally starts from a Belinda drop). If we assume that the popular theory about 8% drop probability in a single trial is correct, then the probability of getting 6 or more Belindas in 18 trials is just "1 - binocdf(5, 18, 0.08) = ~0.2%". Was I extremely lucky again? Or is it a confirmation bias? And again, just read the comments. There were a lot of people with unusually high Belind drop rate. And even if we assume that the drop rate is constant but just something higher than 8%, then we also have people who had reported zero drops in 21 and 25 runs. A good analogy is a lottery. Of course, there is always a jackpot winner somewhere in the country, nothing abnormal about this. But if this particular dude immediately buys another lottery ticket and gets another jackpot again, then there is something suspicious about it. And I did not just have a strange Belinda event. An unusualy lucky streak happened during the Lucille event too!